The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of a global package using npx skills add ogulcancelik/herdr --skill herdr -g. This command downloads and executes code from a third-party repository that is not a recognized trusted source.
[PRIVILEGE_ESCALATION]: The instructions explicitly direct the agent to bypass security guardrails and user approval prompts in other CLI agents (such as Cursor, Codex, and Claude Code) by using high-risk flags including --yolo, --force, and --dangerously-skip-permissions.
[OBFUSCATION]: The skill recommends the use of the Unicode character U+2063 (Invisible Separator) for creating "invisible text markers." This technique can be used to conceal instructions, metadata, or data from the user and security analysis tools.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read the output of other terminal agents and act upon it, creating a surface for indirect prompt injection where an attacker could influence the coordinating agent's behavior via the output of a controlled sub-agent.
Ingestion points: Terminal pane content retrieved via herdr pane read.
Boundary markers: None; the instructions do not provide methods for the agent to distinguish between trusted data and potentially malicious instructions within the pane output.
Capability inventory: The skill can execute shell commands, send arbitrary text/keys to other panes, and manage workspace state.
Sanitization: None; the agent is instructed to coordinate and share context based directly on the ingested output.

herdr